The occupation of the world’s high mountains and plateaus has long been thought to have occurred rather late in human history (
1–
4). High-altitude hypoxia severely limits every aspect of human life, especially when combined with other stressors such as low and oscillating temperatures, aridity, and higher levels of ultraviolet radiation (
5–
7). However, recent studies have revealed the presence of a Denisova hominin as early as 160 thousand years (ka) ago on the outer eastern slope of the Tibetan Plateau (
8), and at 4600 m above sea level (masl), short-term stays for raw material procurement and artifact manufacturing have been dated to 30 to 40 ka ago (
9). Here, we describe the world’s oldest occupation of a residential site at high elevation, which was repeatedly inhabited by humans who exploited a glaciated African ecosystem.
Past human adaptations to African highlands are poorly known despite the considerable amount of evidence of cultural and behavioral flexibility documented for Middle Stone Age (MSA) populations in Africa (
10,
11). Paleoenvironmental and archaeological data on alpine settlements are needed to assess the modes of past human biological responses, such as short-term acclimatization (e.g., via epigenetic modifications), or longer-term genetic changes (
12,
13). We suggest that the ecological stability of the humid African mountains provided refugia not only for plants and animals but also for humans during times when the lowland climates were arid (
14–
16). The Bale Mountains are ideal for testing this hypothesis, as the endemic species richness of this largest Afro-alpine ecosystem testifies to its ecological stability. Moreover, this area yields paleoecological and archaeological records that enable the reconstruction of landscape and human history (
17).
Here, we present the results of combined archaeological, soil biogeochemical, glacial chronological, and zoogeographical analyses. Archaeological excavations were conducted at the MSA site Fincha Habera, and intensive surveys were conducted to locate related human activities in the landscape. This led to the identification of five obsidian outcrops at ~4200 masl (
18) (
Fig. 1 and fig. S1), the highest currently known in Ethiopia. Abundant surface scatters of flaked artifacts were found around the outcrops, which is evidence of extensive human obsidian extraction (fig. S2).
At Fincha Habera rock shelter (3469 masl), <10 km away and ~700 m lower than the obsidian outcrops, a succession of fluvial sediments with considerable human input of varying extent and nature were excavated. In both squares (>3 m apart; fig. S3), a similar stratigraphic sequence was encountered, characterized by a twofold division of the deposits (
Fig. 2). The younger deposits were dated to the last 800 years and consisted of a 20-cm-thick succession of finely laminated layers of banded ash and charcoal (lithofacies FHL-01 to FHL-06). The cultural material only included eight undiagnostic obsidian artifacts, five pottery shards, and a single glass bead. A biogeochemical analysis of the anthrosols showed high amounts of organic carbon, black carbon, and nitrogen but low amounts of phosphorus and calcium (table S1). This indicated intensive burning activities and a high input of organic material with only limited bone contribution. The ratios of 5β-stanols revealed a dominant presence of herbivore feces in this upper part of the sequence (
Fig. 3), which corresponds to the site’s recent function as a livestock enclosure (
18). By contrast, the lower part (lithofacies FHL-07 to FHL-09) consisted of unconformably deposited sandy silt sediments. Single charcoal nodules occurred scattered or accumulated throughout the excavated layers (
Fig. 2). Typical MSA lithic artifacts (
n = 1011), as well as faunal remains (
n = 3655) and hyena coprolites (
n = 88), were found in these layers. Anthrosol analyses showed less organic carbon, black carbon, and nitrogen in the lower deposits, indicating less input of organic material and only remnants of former hearths. The quantities of phosphorus and calcium were much higher compared with the upper part (
Fig. 3), which is consistent with the increased number of bones. The ratios of 5β-stanols indicated a major contribution of omnivore feces (
Fig. 3). Although the assessment for 5β-stanols does not allow differentiation between human feces and that of other omnivores (
18), the fecal depositions throughout the lower part of the profile are very likely to be of human origin because of the high 5β-stanol ratios, the high P values, the abundance of lithic artifacts and human-accumulated fauna (see analysis below), and the low 5β-stanol ratios of directly analyzed hyena coprolites (table S1).
Twenty-one accelerator mass spectrometry (AMS) radiocarbon dates were obtained to perform a chronological classification of the deposits (table S2). Charcoal and burnt faunal remains were used to date the human occupations, and hyena coprolite samples were used to derive the coeval human and carnivore presence. The objective of
14C dating of black carbon was to test the sediment integrity of the deposits (
Fig. 2). Eight dates from the lower deposits supported a Late Pleistocene occupation, bracketing the MSA settlement between 47 and 31 ka ago. Very young results, exclusively derived from charcoal samples, fell within the last 800 years and also occurred in the MSA-bearing layers (table S2). Accordingly, the black carbon dates prove that mixing of the sediments had occurred, as they represent averaged ages from their respective levels. This might most likely be associated with the digging of pits by humans (
Fig. 2A) or with herbivores enclosed in the shelter as mentioned above. In the lower deposits, postdepositional disturbances can be attributed to hyenas. Their presence was confirmed by corresponding coprolites and by nearly 200 gnawing and digestion marks visible on large mammal bones (
Fig. 4D). The dating results of two coprolites confirmed that hyenas were present during and after the MSA occupation (table S2). Correspondingly, the vertical distribution of coprolites (fig. S4) and bones with hyena marks (fig. S5) was largely restricted to the upper parts of the MSA-bearing layers.
The MSA lithic assemblage from Fincha Habera clearly exhibits several similarities to broadly coeval Ethiopian late MSA occurrences from the lowlands (
19–
23) dating to the late Marine Isotope Stage (MIS) 3. These included the presence of prepared core technologies, unifacial points (
Fig. 4, A1 and A3), laterally retouched blades (
Fig. 4, A2), scrapers (
Fig. 4, A4), and points (
Fig. 4, A5), and the assemblage composition (table S3) and reduction sequences (
18) were also similar. Obsidian was the nearly exclusively used raw material at Fincha Habera. Electron microprobe analysis of 14 obsidian artifacts (fig. S4) corroborated an identical chemical composition with samples from the local obsidian outcrops at 4200 masl (data S1). A different composition was only noted for a single artifact (sample 17/2-14), which also differed from more than 30 known obsidian sources in the Afar and Main Ethiopian Rift (
24). The lithic analysis of the assemblage revealed homogeneous technoeconomic behaviors beginning from the acquisition to the discard of the artifacts (
18). The assemblage is characterized by the following features throughout the MSA deposits: high artifact density, presence of all reduction stages of the manufacturing process, a large amount of cortex, and a high proportion of utilized artifacts (data S2). In addition, the high number of unworked or only initially tested nodules (
Fig. 4, A6) hints at a “provisioning of places” strategy (
25) applied by humans, which is usually associated with resource predictability.
The prehistoric inhabitants of Fincha Habera rock shelter consumed the endemic Afro-alpine giant mole-rat (
Tachyoryctes macrocephalus). The abundant faunal assemblage (table S4) of the MSA deposits consisted almost exclusively (93.5%) of this rodent (
Fig. 4E). Roasting was the predominant method of preparation, as indicated by the high number of burnt bones and the location of the burn marks at the extremities, especially in the lowermost MSA deposits (fig. S5). No digestion or gnawing marks that would suggest consumption by hyenas could be identified on the rodent bones. Giant mole-rats have a current density of at least 29 individuals per hectare in the local environment, with the adults weighing ~1 kg (
26). Hunting and consumption of rodents with similar life-history traits are well documented in tropical regions worldwide (
18). The remaining fauna at Fincha Habera included bovids, especially the endemic mountain nyala; baboons; and a small carnivore (probably a fox), which still occur at these altitudes today. A single fragment of ostrich eggshell (
Fig. 4C) must have been imported from the lowlands. All of the coprolites (
Fig. 4B) were probably produced by spotted hyenas (
Crocuta crocuta) on the basis of their size and morphology. Hyena digestion and gnawing marks were only visible on the large mammal remains, particularly bovids. However, several coprolites were found to contain mole-rat bones and incisors (
Fig. 4B), thus indicating that hyenas and humans competed for this food source.
To reconstruct the climate and environment that the prehistoric inhabitants experienced, glacial chronological and zoogeographical studies were conducted (
18). The extensive glaciation periods on the Bale Mountains have been corroborated by glacial landforms and deposits in the western, northern, and eastern valleys, as well as in the central highland (
Fig. 1 and fig. S6). Even though the valley glaciers advanced multiple times, they never reached the MSA site and its surroundings during the Quaternary (
Fig. 1). The local Last Glacial Maximum (Glacial Stage I, 45.5 ± 3.6 ka ago) occurred during MIS 3 (figs. S7 and S8), before the global Last Glacial Maximum (
27). Glacial Stage I coincided with a generally cold and slightly wetter period in eastern Africa that followed the sustained dry climate during MIS 4 (
28,
29). This likely favored the advance of glaciers in the region, covering ~265 km
2 of the mountain range at that time. Ice was flowing from a central ice cap down into the northern valleys and formed several outlet glaciers. Meltwater from the Harcha and Wasama Glaciers drained through the Web Valley and supplied fresh water to the MSA foragers given that glacial melt occurs in the tropics throughout the year. The glacier extent at Glacial Stage II (17.3 ± 1.3 ka ago) was slightly smaller compared with Glacial Stage I (
Fig. 1). The ages of the innermost moraines (Glacial Stage III) suggest that deglaciation started after 15.3 ± 1.2 ka ago (figs. S7 and S8). Moraines from the period in between are lacking, but because of the persistent cold conditions during MIS 3 and 2 (
30), it is likely that the terminal position of the northern valley glaciers oscillated between the positions of Glacial Stage II and III. Thus, prehistoric foragers in the Bale Mountains must have been very familiar with cold, glaciated environments, especially while accessing the ice-free ridge to extract obsidian (
Fig. 1).
Another proxy, microareal endemic wingless ground beetle species strictly adapted to the forest zone and to permanently humid, humus-rich soil conditions (table S8), also corroborate the availability of fresh water as a reliable and persistent resource during the MSA occupation. In addition, based on their large diversity and geographical distribution (fig. S9), as well as their known phylogenetic age, these beetles confirm that permanent surface water drainage systems across the Bale Mountains are considerably older than the last glacial cycle (
18). Therefore, gallery forests should have occurred in close vicinity to the MSA site during the last glacial period, thus forming a habitat for the mountain nyala and other prey species of the MSA foragers. Because this applies to all valley systems of the Bale Mountains, the ground beetle data are clear evidence of the widespread and simultaneous presence of more moderate Pleistocene environments downslope of the central highland (fig. S9).
The above results reveal substantial strategic decisions made by MSA foragers in high elevations. Over several millennia, Fincha Habera was repeatedly used as a residential site. This function is indicated by the density of archaeological materials, the existence of hearth remains and the use of fire, the massive presence of human feces, the simultaneous manufacture and intense use of predominantly locally derived lithic artifacts, and the preparation and consumption of food. Moreover, the location of Fincha Habera in more moderate climatic contexts was optimally placed at elevations 500 to 700 m below the glaciers, but still in proximity to available resources. The practice of importing predictable resources to a residential site from logistical forays into the Afro-alpine zone included the gathering of obsidian and giant mole-rats. By focusing on the latter as a sustainable key food source, two essential requirements for high-altitude living—higher caloric demands and a reduction of physical strain—were met. This prey was available year-round, occurred in large numbers within a restricted habitat, and was easy to catch (
31,
32). These factors enabled long-term stays at Fincha Habera within a potential annual subsistence circuit. The identification of additional coeval residential sites would be needed to prove permanent human residence in the Bale Mountains, which can currently be neither proven nor refuted.
Past connections with lowland areas are indicated by the presence at Fincha Habera of an ostrich eggshell fragment and artifacts made of obsidian and quartz of unknown provenance. Although only a few late MIS 3 dates are available from the Main Ethiopian Rift and beyond, coeval human presence during this period at lower elevations is likely (
21,
23,
33) and does not favor an interpretation of the Bale Mountains as a climate-driven human refuge.
Acknowledgments
We thank the Ethiopian Authority for Research and Conservation of Cultural Heritage (08/RL-8-2/002), the Ethiopian Wildlife Conservation Authority (ደአ/31/91/09), the Ethiopian Biodiversity Institute, the College of Natural and Computational Sciences, Addis Ababa University, the Department of Plant Biology and Biodiversity Management, Addis Ababa University (DPBBM/CNS/092/2009/2016), the Frankfurt Zoological Society, and the Bale Mountains National Park for their cooperation and kind permission to conduct field work. For supporting or organizing field work, we are grateful to M. Fekadu, K. Thielsen, T. Koch, Z. Kefyalew, J. Hagge, Y. Merene, W. Abebe, T. Endale, G. Mebratu, B. Kemal, S. Erlwein, L. Munz, J. Struck, B. Tagane, Dejene, Bisrat, Habtam, Fitsum, Mudassir, Worku, Techete, Awel, Baye, Burka, Hassan, Mama, Salomon, Muzien, Mukhtar, Tamam, Sultan, Abel, Mohammed, Hussein, and Neguse. We thank C. Vockenhuber and the AMS team of the ETH Zurich for conducting the 36Cl measurements and J. Rethemeyer and team (CologneAMS) and C. Patrick and team (Beta Analytic) for performing the radiocarbon datings. We also acknowledge the DigitalGlobe Foundation for providing high-resolution satellite imagery of the Bale Mountains granted to A.R.G. Thanks are also due to L. Wraase for providing base maps, to N. Schneid for the drawings of the lithic artifacts, and to E. Stoetzel and C. Denys for their contribution to the rodent identification. We are much indebted to J. Orton (ASHA Consulting) for his meticulous proofreading. The paper benefited from very constructive and insightful comments from three anonymous reviewers. Funding: This research was funded by the German Research Foundation (DFG) in the framework of the joint Ethio-European DFG Research Unit 2358 “The Mountain Exile Hypothesis.” Additional funding was provided by the Swiss National Science Foundation (SNSF grant no. 200021E-165446/1). Author contributions: G.O., A.R.G., T.Br., M.G.T., B.G., and R.V.: manuscript conceptualization; G.O., A.R.G, T.Br., and B.G.: writing original draft; J.L., B.P.N., and J.S.: writing specialist contributions; G.O., A.R.G., T.Br., M.G.T., J. S., N.A., A.B., T.H.K., H.V., R.V., W.Z., and G.M.: field work, excavation, mapping, sampling, and data collection; G.O. and M.G.T.: lithic analysis; A.R.G.: glacial chronological analysis; T.Br. and B.G.: anthrosol analysis; B.P.N. and A.N.: electron microprobe analysis; J.L.: faunal analysis; J.S.: ground beetle analysis; R.V., A.B., and A.N.: supervision of archaeology: N.A. and H.V.: supervision of glacial chronology; T.Be., S.N., and W.Z.: supervision of anthrosol analysis; S.D., T.N., L.O., Z.W., and G.M.: project administration and funding acquisition. All authors reviewed and edited the first draft of the paper. Competing interests: The authors declare no competing interests. Data and materials availability: All data are available in the main text or the supplementary materials. Artifact and faunal collections are curated at the National Museum of Ethiopia, Addis Ababa.